← All articles

Optimizely Supply-Chain Breach Hits CNN, BBC, Zoom and 178 Sites With Card Skimmers and iOS Exploit Kits

A third-party script that most security teams never audit just became the delivery vehicle for one of the broadest web-skimming campaigns of 2026. Attackers compromised Optimizely Web Experimentation — an A/B testing and personalization platform embedded on thousands of corporate sites — and used it to push malicious JavaScript into the Project JavaScript field that runs on every page of affected customer projects. Visitors to at least 178 organizations, including CNN, the BBC, Zoom, Salesforce, GitLab, Fox News, HSBC, and IKEA, were served the payload starting November 16, 2026.

This is the kind of attack I spent years hunting for at Deloitte: not a flaw in the target's own code, but in a vendor script every legal and procurement team signed off on without anyone from security reviewing what it actually executes in the browser. That's exactly why it worked at this scale.

What happened

Optimizely Web Experimentation lets customers inject custom JavaScript ("Project JavaScript") that runs on every page where the platform's snippet loads. An attacker gained the ability to write to that field across a large number of customer projects and inserted a loader that pulled a second-stage payload, track.js, from attacker-controlled infrastructure at 1688360[.]cc.

Report URI, the security firm that first caught the activity, timestamped a clean snapshot of its own Optimizely file at 02:21 UTC on November 16. By 04:21 UTC the same file had been modified to include the malicious loader. Four hours later, at 08:28 UTC, the payload was observed executing on bbc.com. Optimizely posted an incident notice at 01:02 UTC on November 17 confirming the vulnerability in Experimentation had been exploited and that malicious JavaScript had been added to customer projects. The company says the flaw is now fixed and the injected code removed.

Separately, researchers at Hudson Rock and the independent analyst behind amirulcyber.com found the same loader pattern injected into other third-party scripts beyond Optimizely — Kissmetrics' i.js, StorifyMe's storifyme-elements.min.js, and minute-ly's loader.js — all pointing at overlapping infrastructure. That suggests either a shared compromise point across multiple SaaS vendors or an attacker deliberately diversifying delivery channels to maximize reach and survive any single vendor's remediation.

Technical breakdown of the payload

track.js is a multi-purpose in-browser implant with three distinct capabilities:

1. Cryptocurrency address swapping. The script scans page text, form fields, hyperlinks, the clipboard, and outgoing network requests for cryptocurrency wallet addresses across Bitcoin, Ethereum, Tron, TON, Solana, Litecoin, Bitcoin Cash, and Dogecoin formats, then silently replaces them with attacker-controlled addresses. A user copy-pasting a BTC address into a payment flow, or a site rendering an address server-side, would have it swapped client-side before it reaches the user's wallet software — a browser-native form of address-poisoning that doesn't require compromising the wallet or the exchange at all.

2. Card skimming. Added at 13:43 UTC on November 16, the skimmer hooks form submissions and intercepted fetch/XMLHttpRequest calls, including JSON bodies and FormData objects, parsing nested structures for anything that passes a Luhn checksum plus a plausible card-brand prefix, a valid MM/YY expiry, and a 3- or 4-digit CVV. Matches are exfiltrated to 1688360[.]cc/card.php. Report URI notes that card data entered into separately hosted, PCI-scope-isolated payment iframes (Stripe, Adyen, and similar) was out of reach — a reminder that iframe isolation for payment fields is one of the few defenses that actually held here.

3. Reconnaissance and exfiltration. The script also captured cookies and page screenshots (via a bundled html2canvas library served from the attacker's own infrastructure) and sent them to 1688360[.]cc/hit.php along with the page URL and any original (pre-swap) wallet addresses it found.

4. iOS exploit delivery. On iPhone, iPad, and iPod visitors specifically, the script loaded boot.js from two .icu domains (5vobauua3ojpklx[.]icu and eil786npto9u5bq[.]icu). That script fingerprinted the iOS version and opened a hidden iframe into one of two exploit kits identified by researchers as Coruna (targeting iOS 13.0–17.2.1) and DarkSword (targeting iOS 18.4–18.7). For iOS 26 devices, the chain currently only collects device telemetry rather than attempting exploitation — likely because no working exploit for the latest major version has been weaponized yet. The kits' own internal comments mark iOS 18.7.3+ and iOS 26.3+ as patched, which is a useful detection signal in itself: it tells you exactly which versions the attacker believed were still vulnerable at the time of the campaign. Report URI reports no confirmed iPhone takeovers, but the presence of a working two-tier exploit chain means the operators were after more than payment data from at least some visitors.

Who is affected

Confirmed or reported affected properties span media (CNN, BBC, Fox News), collaboration and SaaS (Zoom, Salesforce, GitLab, Smartsheet, Hootsuite), finance and insurance (HSBC Insurance Singapore, HSBC UAE, Mutual of Omaha), retail (IKEA), telecom (Virgin Media O2), and others including Tricentis, Stamps.com, and Optimizely's own corporate site. The common thread isn't industry — it's that all of them loaded Optimizely's (or, in a few cases, Kissmetrics', StorifyMe's, or minute-ly's) script unmodified and trusted it by default, the same trust model that makes every third-party tag on your site a de facto supply-chain dependency.

There is no public CVE identifier for the underlying Optimizely platform flaw as of this writing, and the company has not disclosed the exact mechanism that let the attacker write to customer Project JavaScript fields — whether that was a compromised internal credential, an API authorization bypass, or something else. Given the breadth and speed of the injection across 178+ projects, an automated, programmatic compromise of a privileged API or internal deployment path is far more likely than manual per-customer tampering.

Indicators of compromise

Security teams should block or alert on the following, drawn from Report URI, Hudson Rock, and independent researcher analysis:

  • Domains: 1688360[.]cc, 5vobauua3ojpklx[.]icu, eil786npto9u5bq[.]icu, webshell[.]lol, ios-hack[.]icu, l7bcsnwc7nzr3uo[.]icu, je4ft3gn8vxudum[.]icu
  • URL paths: /track.js, /card.php, /hit.php, /html2canvas.min.js, /boot.js, /px.gif
  • SHA-256 of the primary payload track.js: f7b84b9a94f749f8ced3f0112d0a66894e8bbc3b4aa5592021d1b8976a9bad40
  • Attacker-controlled cryptocurrency addresses used for substitution (ETH 0xC4026b06D82b0F0B363eda0642D61CfbDc17884e, BTC 19g9rZcPeiz8rUvyWpTRyXXhPnjsWrc3kW, among others documented by researchers)

Remediation and detection

If your organization uses Optimizely Web Experimentation, Kissmetrics, StorifyMe, or minute-ly:

  1. Verify current snippet integrity. Pull the live Project JavaScript / custom code field for every project today and diff it against a known-good backup or version control history, not just today's vendor-reported "clean" state.
  2. Rotate anything exposed. If the skimmer was live on a payment-adjacent page, assume card data submitted through non-iframed forms during the exposure window is compromised. Notify your payment processor and consider PCI incident disclosure obligations.
  3. Block the IOC list above at your web proxy, CDN, and EDR layer immediately — these domains were registered within the last 48 hours and have no legitimate traffic.
  4. Enforce Subresource Integrity (SRI) and CSP on every third-party script tag, including vendor-hosted ones like Optimizely's. SRI hashes would have broken this attack the moment the vendor file changed, at the cost of needing a redeploy whenever the vendor legitimately updates the script — a tradeoff worth making for anything running on every page.
  5. Isolate payment forms in vendor-hosted iframes (Stripe Elements, Adyen, etc.) rather than raw HTML forms. This is the one control that demonstrably limited damage in this incident.
  6. Patch iOS. Update to iOS 18.7.3 or iOS 26.3 or later; both the Coruna and DarkSword exploit kits treat these versions as unexploitable per the attackers' own code comments.
  7. Audit who can write to your SaaS vendors' customer-facing config. This incident is a reminder that "no-code" personalization and A/B testing tools are effectively unreviewed JavaScript injection points into production — treat vendor API tokens and admin accounts for these platforms with the same rigor as production deploy credentials.

Frequently asked questions

Is there a CVE for the Optimizely vulnerability?

Not publicly disclosed as of this writing. Optimizely has confirmed the incident and says the underlying flaw in Experimentation has been fixed, but has not released a CVE identifier, technical root cause, or scope of affected projects.

Was my payment card data stolen if I visited one of the listed sites?

Only if you submitted card details through a form directly hosted on the affected page's own domain during the exposure window (roughly November 16, 13:43 UTC onward until each project was cleaned). Cards entered into separately hosted payment iframes, such as Stripe Elements or Adyen's hosted fields, were not accessible to this skimmer.

Does this affect iPhones that were just browsing, not entering payment info?

Potentially, yes. The iOS exploit delivery component triggered independently of any form submission, simply by loading a page with the poisoned script on an affected iPhone or iPad. Report URI found no confirmed device takeovers, but the exploit chains for iOS 13–17 (Coruna) and iOS 18.4–18.7 (DarkSword) were live and functional. Update to iOS 18.7.3+ or 26.3+ immediately regardless of whether you believe you were exposed.

How did this spread beyond Optimizely to Kissmetrics, StorifyMe, and minute-ly?

Researchers found the same loader infrastructure embedded in scripts from those three additional vendors. It's not yet confirmed whether this reflects a single shared compromise (e.g., a common CDN, build pipeline, or reseller) or independent exploitation of each vendor using similar techniques. Treat all four as compromised vendors until each publishes its own root-cause disclosure.

What should I tell my security team right now?

Pull every third-party script your production sites load — not just Optimizely — and check it against a known-good copy today. Deploy SRI hashes or a strict CSP script-src allowlist for anything you can't fully control. This incident is a live demonstration of why "it's just a vendor's A/B testing tag" is not a reason to skip security review.

Find what attackers can see, before they do.

AI pentesting agents test your web app, API and cloud, and every finding comes with evidence and a fix. First report in about 10 minutes.

Stay ahead of attackers.

New vulnerabilities, breach write-ups and practical fixes. A few emails a month.